摘要
采用某股份制银行的698家贷款企业样本,基于粗糙集-Elman神经网络集成构建了贷款企业五级分类评估模型.该模型首先应用粗糙集理论约简出重要指标体系,然后将训练样本送入Elman神经网络进行学习和训练,进而对检验样本的风险等级进行判别.结果表明,与传统的logistic回归模型相比,粗糙集-神经网络系统对检验样本预测精度更高,是一种更为有效和实用的分类方法,为我国商业银行五级分类管理提供一个新的方法.
Taking 698 loaning enterprises in a stock commercial bank as samples, this paper proposes integrating five-category model of rough sets and neural network. Firstly, the attributes are reduced using rough set and Elman neural network is trained with training samples , then the model is used to evaluate risk grade of testing samples. Empirical results shows that, comparing with logistic model, integration model of rough sets and neural network is an efficient and practical tool to evaluate credit risk of testing samples, and supplies commercial bank a new method in five-category management.
出处
《系统工程理论与实践》
EI
CSCD
北大核心
2008年第1期40-45,55,共7页
Systems Engineering-Theory & Practice
基金
国家自然科学基金项目“信用风险评估理论、方法及其应用”(70171005)
